Paa improve

[Johnson-Wang]

Continue over #3803
Multiple choices of covariance_type are now supported.

Changing the covariance_type from 'full' to 'diag' does improve the training speed and yield the same performance. Yet the acceleration rate is around 10%-20%.

[CLAassistant]

Thank you for your submission! We really appreciate it. Like many open source projects, we ask that you all sign our Contributor License Agreement before we can accept your contribution.
2 out of 3 committers have signed the CLA.

✅ rydenisbak
✅ Johnson-Wang
❌ d.rybalchenko

---

d.rybalchenko seems not to be a GitHub user. You need a GitHub account to be able to sign the CLA. If you have already a GitHub account, please add the email address used for this commit to your account.
_{You have signed the CLA already but the status is still pending? Let us recheck it.}

[codecov]

Codecov Report

Merging #3985 into master will increase coverage by 0.12%.
The diff coverage is 64.28%.

@@            Coverage Diff             @@
##           master    #3985      +/-   ##
==========================================
+ Coverage   62.17%   62.30%   +0.12%     
==========================================
  Files         221      221              
  Lines       16314    16321       +7     
  Branches     2759     2762       +3     
==========================================
+ Hits        10144    10169      +25     
+ Misses       5678     5657      -21     
- Partials      492      495       +3     

Flag	Coverage Δ
#unittests	62.30% <64.28%> (+0.12%)	⬆️

Flags with carried forward coverage won't be shown. Click here to find out more.

Impacted Files	Coverage Δ
mmdet/models/dense_heads/paa_head.py	89.79% <64.28%> (-1.81%)	⬇️
mmdet/models/roi_heads/mask_heads/maskiou_head.py	100.00% <0.00%> (+5.37%)	⬆️
mmdet/models/roi_heads/mask_scoring_roi_head.py	87.50% <0.00%> (+32.14%)	⬆️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update 5d59215...2740161. Read the comment docs.